p10436 and p10437

Coming to -advmethods today are two interesting WUs. They are classic WUs that use the new a4 core. The a4 core will allow a WU to be run in single or multi-threaded mode, so if you're running v6.34 of the client with the -smp flag and advanced methods set, you will start seeing more and more of WUs of this type. Production is slightly higher than average a3 ppd. Who knows, it may turn out to be a setting that avoids p670x.

ChasR said:

Coming to -advmethods today are two interesting WUs. They are classic WUs that use the new a4 core. The a4 core will allow a WU to be run in single or multi-threaded mode, so if you're running v6.34 of the client with the -smp flag and advanced methods set, you will start seeing more and more of WUs of this type. Production is slightly higher than average a3 ppd. Who knows, it may turn out to be a setting that avoids p670x.

Click to expand...

Thanks for the news. So does this mean that our current configs are still good?

So I'll add the new 634 client to my smp rig and check that advmethods is set. Correct?

I had left this client at 630.

EDIT: adv is set... D/L 634....

What do we know about a 7148? I switched to the 634 client added "yes" advmethods in advanced settings and pulled one of these! I hit a personal smp high on it 14k as opposed to 11500k or so at its previous best. Was it the client or the advmethods flag?

Thanks for this great tip Chas, I recommend it to any one doing smp on an i7.

Ya gotta love free PPD!

For that particular WU, it was the flag.

Q9550 @ 3.6 in a XP/VM/Ubuntu with 2 x 8800 STS (G92)
Project ID: 7154
Core: GRO-A3
Credit: 585
Frames: 100



Name: Val-VM 10.10
Path: \\VAL-UBUNTU\fah\
Number of Frames Observed: 13

Min. Time / Frame : 00:04:27 - 12,240 PPD
Avg. Time / Frame : 00:04:32 - 11,904 PPD
Cur. Time / Frame : 00:04:39 - 11,420 PPD
R3F. Time / Frame : 00:04:45 - 11,077 PPD
All Time / Frame : 00:04:32 - 11,843 PPD
Eff. Time / Frame : 00:04:53 - 10,646 PPD

Anyone seen any of these WUs yet?

Just a note, HFM will assume the client type for these WUs is still Standard; even if run with the -smp flag. The change by Stanford breaks a long running scheme of a singular Core == Client relationship.

To be honest I've considered doing away with the Standard vs. SMP distinction all together and simply referring to the client type as CPU or GPU.

Edit, Q6600 @ 3.0 and 3.2 Ghz

Project ID: 10437
Core: GRO-A4
Credit: 311
Frames: 100


Name: Nikki SMP
Path: \\Cpr-ws04\FAH SMP\
Number of Frames Observed: 100

Min. Time / Frame : 00:03:54 - 6,349 PPD
Avg. Time / Frame : 00:03:57 - 6,229 PPD


Name: WS-16 Win SMP
Path: \\Cprcc-ws16\Q5 FAH\SMP2\
Number of Frames Observed: 200

Min. Time / Frame : 00:03:33 - 7,311 PPD
Avg. Time / Frame : 00:03:47 - 6,645 PPD

Both of these are running with a g92 gpu.

Project ID: 10437
Core: GRO-A4
Credit: 311
Frames: 100

F34 Q9550 at 3.7 with GTS250 gpu client.
Name: F34-FAH1-P5B-Q9550-XP32
Number of Frames Observed: 100
Min. Time / Frame : 00:02:57 - 9,651 PPD
Avg. Time / Frame : 00:03:02 - 9,256 PPD

F42 Q9550 at 3.8 with two 9600GSO clients.
Name: F42-FAH1-P5KEw-Q9550-XP32
Number of Frames Observed: 203
Min. Time / Frame : 00:02:51 - 10,164 PPD
Avg. Time / Frame : 00:02:53 - 9,988 PPD